The field of speech translation and sentiment analysis is moving towards more fine-grained modeling of speech features and semantic spaces. Researchers are exploring new approaches to mitigate the challenges of data scarcity and linguistic diversity, such as leveraging large language models, mixture of experts, and synthetic parallel data. Notable papers in this area include: Towards Fine-Grained Code-Switch Speech Translation with Semantic Space Alignment, which proposes a novel approach to code-switch speech translation using a mixture of experts speech projector. TEDxTN: A Three-way Speech Translation Corpus for Code-Switched Tunisian Arabic - English, which introduces a new publicly available dataset for speech translation. Improving Direct Persian-English Speech-to-Speech Translation with Discrete Units and Synthetic Parallel Data, which presents a direct speech-to-speech translation system using discrete units and synthetic parallel data. AHaSIS: Shared Task on Sentiment Analysis for Arabic Dialects, which focuses on sentiment detection in Arabic dialects and provides a multi-dialect dataset. MAPROC at AHaSIS Shared Task: Few-Shot and Sentence Transformer for Sentiment Analysis of Arabic Hotel Reviews, which employs few-shot learning for sentiment analysis of Arabic dialects.